Research map / Point clouds / Point cloud classification
Adversarially Robust 3D Point Cloud Recognition Using Self-Supervisions
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adversarial jigsaw pre-training before adversarial fine-tuning (ablation) | DGCNN AT · ModelNet40 | robust accuracy: 62.0 -> 65.8 | Helped |
| add adversarial jigsaw pre-training before adversarial fine-tuning (ablation) | PointNet AT · ModelNet40 | robust accuracy: 37.9 -> 52.1 | Helped |
| add adversarial jigsaw pre-training before adversarial fine-tuning (ablation) | PCT AT · ModelNet40 | robust accuracy: 49.1 -> 51.3 | Helped |
| add adversarial jigsaw pre-training (ablation) | DGCNN AT · ScanObjectNN | robust accuracy: 30.9 -> 41.3 | Helped |
| ensemble fine-tuned models from different SSL tasks (max logits) (ablation) | APF models · ModelNet40 | robust accuracy: 62.0 -> 69.4 | Helped |
| use adversarial joint training with SSL instead of pre-training (ablation) | PCT AT · ModelNet40 | robust accuracy: 49.1 -> 44.3 | Hurt |
| add adversarial joint training with rotation (eta=6) (ablation) | PointNet AT · ModelNet40 | robust accuracy: 37.9 -> 45.0 | Helped |
| add joint training with jigsaw (k=3) (ablation) | DGCNN AT · ModelNet40 | robust accuracy: 62.0 -> 62.3 | No clear effect |
| add jigsaw pre-training under point dropping attack (ablation) | DGCNN AT · ScanObjectNN | robust accuracy: 77.1 -> 80.7 | Helped |
| add adversarial jigsaw pre-training under point adding attack (ablation) | DGCNN AT · ScanObjectNN | robust accuracy: 63.5 -> 68.2 | Helped |
| add jigsaw pre-training for part segmentation (ablation) | DGCNN AT · ShapeNetPart | robust mIoU: 69.5 -> 70.3 | Helped |
| select piecewise learning-rate decay, remove T-Nets, replace FPS with EdgeConv (ablation) | AT baselines · ModelNet40 | robust accuracy: 44.2 -> 49.1 | Helped |
Closest papers by meaning
- Explicitly Perceiving and Preserving the Local Geometric Structures for 3D Point Cloud Attack (AAAI 2024 · similarity 0.88)
- Curvature-Invariant Adversarial Attacks for 3D Point Clouds (AAAI 2024 · similarity 0.88)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.87)
- CausalPC: Improving the Robustness of Point Cloud Classification by Causal Effect Identification (CVPR 2024 · similarity 0.87)
- Improving Robustness of 3D Point Cloud Recognition from a Fourier Perspective (NeurIPS 2024 · similarity 0.87)
- A Unified Approach to Interpreting Self-supervised Pre-training Methods for 3D Point Clouds via Interactions (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.